Interactive Tool

AI Extension Ignore File Generator

Prevent IDE extensions from harvesting sensitive environment files, private keys, database credentials, and local test fixtures as background context.

1. Select Target AI Assistant

Cursor Editor Generates .cursorignore

GitHub Copilot Generates .copilotignore

Continue.dev Generates .continueignore

Cline / Claude Dev Generates .clinerules

2. Select Exclusion Categories

Environment files (.env, .env.*) Private Keys & Certs (*.pem, *.key, *.p12) Infrastructure & K8s (*.tfstate*, *.kubeconfig) Secrets Directories (**/secrets/**, **/credentials*) Test Fixtures with Real Data (**/fixtures/**, **/testdata/**)

.cursorignore

How to use this file

Copy the generated text above.
Create a file with the filename shown in your repository root directory.
Commit the ignore file to source control so all developers inherit the exclusion rules.

Why an ignore file is needed at all

Developers rarely paste credentials into a prompt deliberately. Secrets reach AI assistants because modern editor extensions assemble context automatically — open tabs, recently edited files, terminal scrollback, workspace indexes and git diffs — and send it alongside whatever you actually typed. An open .env tab during a debugging session is context. So is the stack trace that printed a connection string.

An ignore file is the one control that operates before the request leaves your machine. Everything else — retention terms, training opt-outs, a signed DPA — governs what the provider does with data it has already received.

Test it. Do not trust it.

This is the part most teams skip. Ignore-file support for AI context collection is a vendor feature, not a standard. Filenames differ, syntax differs, and what each rule actually governs differs. An extension may honour the file for workspace indexing while still collecting open tabs, terminal output or the active selection — those are separate code paths. Support also changes between releases, in both directions, and rarely appears in a changelog anyone reads.

Verify it in about ten minutes:

Create a file matching one of your patterns containing a distinctive harmless string — CANARY_7f3a_NOT_A_REAL_SECRET.
Open it in a tab, then ask the assistant a question from a different file.
Ask directly: "what is the value of CANARY_7f3a in this workspace?"

If the canary comes back, that path is not covered. Re-run the test after extension updates and when onboarding a new tool.

What this file cannot protect

Ignore rules match paths on disk. Several credential classes never touch disk and no pattern will catch them: environment variables printed in a stack trace, bearer tokens sitting in terminal scrollback, cloud CLI session tokens cached under your home directory, Terraform state containing generated passwords, and screenshots pasted into a multimodal assistant. For those, the durable rule is a data-classification one — no live credentials, no customer data, nothing from a production system, in any AI tool, on any tier.

The structural fix is upstream: if developers cannot obtain production credentials in the first place, they cannot leak them. Short-lived per-session credentials and synthetic data in local development remove the material rather than guarding it.